Article, Emergency Medicine

Electronic medical record-based interventions to encourage opioid prescribing best practices in the emergency department

a b s t r a c t

Objective: Overdose from opioids has reached epidemic proportions. Large healthcare systems can utilize existing technology to encourage responsible Opioid prescribing practices. Our study measured the effects of using the electronic medical record (EMR) with direct clinician feedback to standardize opioid prescribing practices within a large healthcare system.

Methods: This retrospective multicenter study compared a 12 month pre- and post-intervention in 14 emergency departments after four interventions utilizing the EMR were implemented: (1) deleting clinician preference lists,

(2) defaulting dose, frequency, and quantity, (3) standardizing formulary to encourage best practices, and

(4) creating dashboards for clinician review with current opioid prescribing practices. Outlying clinicians received feedback through email and direct counseling. Total number of Opioid prescriptions per 100 discharges pre- and post-intervention were recorded as primary outcome. Secondary outcomes included number of prescriptions per 100 discharges/clinician exceeding 3-day supply (defined as 12 tablets), number exceeding 30 Morphine equivalent daily dose (MEDD)/day, and number of non-formulary prescriptions.

Results: There were N700,000 discharges during pre- and post-intervention periods. Percentage of total number Opioid prescriptions per 100 discharges decreased from 14.4% to 7.4%, a 7.0% absolute reduction, (95% CI,6.9%-

7.2%). There was a 5.9% to 0.7% reduction in prescriptions exceeding 3-days, (95% CI, 5.1%-5.3%), a 4.3% to 0.3% reduction in prescriptions exceeding 30 MEDD, (95% CI, 3.9%-4.0%), and a 0.3% to 0.1% reduction in non- formulary prescriptions, (95% CI, 0.2%-0.3%).

Conclusions: A multi modal approach using EMR interventions which provide real time data and direct feedback to clinicians can facilitate appropriate opioid prescribing.

(C) 2019

Introduction

The overuse of Prescription opioids and increasing rate of opioid- related deaths has led the Centers of Disease Control and Prevention to declare opioid overdose a United States epidemic [1]. Despite reductions in opioids prescribing from 2010 to 2015, the amount of opioids being prescribed in 2015 remained as much as three times higher than 1999 [2]. Large healthcare systems are investigating interventions to decrease opioid overprescribing in the inpatient and outpatient arena [3,4].

In 2016, the US Department of Health and Human Services reported more than 63,600 drug overdose deaths in the United States [5]. Emergency departments (EDs) are one area in the clinical arena where patients present for pain control. While prescriptions from the Emergency department (ED) have increased over the past 20 years, other Clinical sites, such as office-based practices and inpatient settings

have seen significant upward trends as well [6,7]. In the emergency medicine literature, Hoppe et al. demonstrated that opioid naive patients prescribed opioids for acute pain in the ED were at increased risk of opioid use at 1 year [8]. Inappropriate prescribing methods from inpatient and outpatient areas have led EDs to become a figurative revolving door where a patient can be discharged with an opioid prescription, and then be received back for treatment of overdose [9-11]. All of these practice settings have the opportunity to benefit from standardized opioid prescribing practices.

National efforts have focused on supporting a more balanced approach to opioid prescribing by implementing state regulatory controls on initial opioid prescribing, creation of databases to track opioid prescriptions for patients, and risk calculators for opioid exposure. Studies have examined electronic medical record (EMR) changes and opioid prescribing policies. One study examined opioid prescribing in the EMR and found that when removing a defaulted

https://doi.org/10.1016/j.ajem.2019.158500

0735-6757/(C) 2019

number of doses for opioid prescriptions, it resulted in providers prescribing a higher quantity than when the EMR gave a set number of pills [12]. Improvement in opioid prescribing has also been demonstrated when implementing an ED Opioid Prescribing Policy to reduce total number of prescriptions, although one study showed similar opioid equivalents as measured by morphine equivalent daily dose (MEDD) even with a decrease in the number of prescriptions [6, 13]. Lastly, providing physicians with their prescribing data has also been shown to have some improvement. Michael et al. found that clinicians were many times inaccurate when asked about self- perception of opioid prescribing practices and Burton et al. found that ED physician prescribing variability can be decreased by sharing their prescribing data amongst other physicians in the group [14,15].

Utilizing multiple strategies from the studies above, the goal of our study was to measure the effects of using the EMR to limit opioid prescriptions and provide direct physician feedback in multiple EDs within a large healthcare system using a multi-faceted approach. By utilizing the EMR and dashboards, our 4-tiered approach consisted of

(1) removing opioids from clinician personal preference lists in the EMR to ensure that physicians and advanced practice providers are adhering to current opioid prescribing practices (2) defaulting a dose, frequency, and quantity of doses in opioid prescriptions to below state-mandated thresholds, (3) standardizing the formulary for more appropriate opioid prescribing practices, and (4) creating dashboards for clinicians to review and to compare their opioid prescribing practices. Our primary outcome was to compare the total number of opioid prescriptions per 100 discharges pre- and post-intervention across the healthcare system. Our secondary outcome was to measure individual clinician practices as measured by clinician rates of opioid prescriptions at discharge, prescriptions exceeding 3-day supply , prescriptions exceeding 30 MEDD, and non-formulary prescriptions.

Methods

Study design

This was a retrospective multicenter study with intervention in 14 EDs across a large integrated healthcare system. All patients who were discharged from the ED during the study time period, 12 month pre- intervention and 12 months post-intervention, were included. This study was approved by the Institutional Review Board of the healthcare system.

Setting

All EDs were included in the analysis. The EDs in the healthcare system were comprised of an urban academic quaternary care center, 5 urban community EDs, 3 suburban community EDs, 2 pediatric EDs, and 3 free-standing EDs with a total annual census of N506,500 patient encounters/year. Pre-intervention data was collected for one year from July 1, 2016 to July 1, 2017. A four part intervention was implemented in stages from July 2, 2017 and December 14, 2017, as this was the length of time required to make changes at all hospitals in our healthcare system. After the implementation period was complete, post- intervention data was collected for one year from December 15, 2017 to December 15, 2018. All sites used the same EMR system (EPIC; Verona, WI) allowing for common accessibility and data acquisition across the system. All clinicians within the system were included in the data analysis which was defined as physician or physician assistant.

Intervention

In alignment with state guidelines, opioid prescription policies were implemented into the EMR as follows: a maximum of 30 mg of MEDD was mandated, all controlled substances required an associated diagnosis on the written prescription, and documentation of an

exception to the acute pain prescribing rules were required either on the prescription or within the EMR. State legislation limited acute pain episodes to maximum of a seven-day supply of opioids in adults, and a five-day supply in pediatrics. EDs were encouraged to limit prescriptions to a three-day course, which we adopted for all patient populations as a maximum of 12 tablets or doses per prescription. Lastly, a closed ED “formulary” was developed within the EMR to default to prescribing of immediate-release formulations of opioids at the lowest tablet strengths commercially available, (Table 1).

To assist prescribers in complying with restrictions, a number of Clinical decision support tools (CDST) were implemented in the EMR to interrupt prescribing workflow. First, an active alert limited the prescribers if attempting to order an opioid at discharge that was not included in the aforementioned ED opioid formulary. Prescribers could override the alert but were required to provide a reason. Real-time calculation of MEDD and the ordered day’s supply would alert the clinician if the number of tabs ordered exceeded the state-mandated 30 MEDD or the ED recommended three day supply limit. Additionally, along with active alerting, non-interruptive CDST methods were employed to limit supplies by minimizing defaulting doses, frequencies, and dispense quantities in the ED opioid discharge orders. Prior opioid orders saved to prescribers’ personal Medication lists were deleted.

As part of the intervention, data was collected on reasons for

clinician override of active alerts, overall prescribing preferences of clinicians for acute complaints seen in the ED, and data on clinician outliers. Clinicians were then given access and direct feedback regarding their prescribing practices through use of an electronic dashboard. Direct feedback was a two-part process. First, monthly dashboards accessible to all clinicians tracked each clinician’s number of total opioid prescriptions, number of prescriptions exceeding a three-day supply (defined as 12 doses), number of prescriptions exceeding 30 MEDD, and the number of non-formulary prescriptions. Clinicians were sent an email with a direct link to the dashboard once per month where they could review their own metrics and compare to the healthcare system as a whole. Clinicians who were outliers for 2 months received direct, active feedback and coaching in person regarding their Prescribing habits. The medical director from the site where the clinician primarily worked made contact with the physician or physician assistant for a meeting to discuss barriers to more appropriate opioid prescribing. The clinician was then closely monitored for improvement by the site director.

Data collection

Data was collected on all ED-basED prescribers in the system, defined as physicians and physician assistants. Individual patient data such as demographics and visit diagnoses were not collected as all patients presenting to the ED who were discharged with an opioid prescription were included in the analysis. Pre-intervention was defined as July 1, 2016 to July 1, 2017 and post-intervention was defined as December 15, 2017 to December 15, 2018. The total number of prescriptions per 100 discharges pre- and post-intervention throughout the healthcare system was calculated. Additionally, clinician level data which included the rates of discharge encounters with opioid prescriptions, rates exceeding 3 days’ supply, rates exceeding 30 MEDD, and rates of non-

Table 1

Formulary for recommended opioid prescriptions.

Oxycodone 5 mg tablets Hydrocodone 5 mg tablets Tramadol 50 mg tablets

Acetaminophen 300 mg/codeine 30 mg tablets Morphine IR 15 mg tablets

Combination products containing acetaminophen, ibuprofen, aspirin also approved

Equivalent liquid formulations also approved

formulary prescriptions were analyzed. Physicians or physician assistants were excluded from the clinician portion of the study if they did not have a minimum of 25 discharges in both the pre-intervention and post-intervention period, however their data was included in overall system level analysis. Opioid prescribing data on each physician and physician assistant was published monthly to a dashboard starting in September 2017.

Statistical analysis was conducted using SAS(R) software (SAS Institute, Inc., Cary, NC). Comparisons of rates between pre- and post- intervention periods were performed using Wald z-tests and confidence intervals for proportions. Waterfall plots were used to graph the pre- and post-intervention periods. For each metric, clinicians were arranged from left to right in descending order, based on pre- Intervention rates; blue bars designate a post-intervention decrease, while red bars show an increase in outcome.

Results

506,779 patients were seen in the pre and 507,877 patients were seen in the post-intervention period. There were N700,000 discharge encounters during the 12 month pre- and post-intervention period. Of those, 369,794 were discharged during the pre-intervention period and 363,691 during the post-intervention period. 361 clinicians met inclusion criteria. 53,337 patients received an opioid prescription prior to the intervention and 26,872 patients received an opioid prescription upon discharge in the post-intervention period. There was a significant decrease in the percentage of opioid prescriptions written upon ED discharge between the pre-implementation and post-implementation period from 14.4% to 7.4%, (95% CI, 6.9%-7.2%), a 51% decrease, (Fig. 1,

Table 2).

In addition, the percentage of prescriptions exceeding a 3-day supply decreased from 5.9% to 0.7%, (95% CI, 5.1%-5.3%). The percentage of prescriptions exceeding a 30 MEDD per day were reduced from 4.3% to 0.3%, (95% CI, 3.9%-4.0%), and the percentage of prescriptions written that were non-formulary decreased from 0.3% to 0.1%, (95% CI, 0.2%-

0.3%), (Fig. 1, Table 2).

Fig. 2 demonstrates individual provider differences between pre- and post-intervention period on percentage of opioid prescriptions written, percentage exceeding 3-day supply as defined as 12 tablets, percentage exceeding 30 MEDD per day, and percentage of prescription written that were non-formulary. In these waterfall plots, every clinician who met criteria of 25 discharges within the pre- and post- intervention period was included. Each bar represents an individual clinician and their overall change in prescribing practices.

Fig. 1. Encounter-Level Changes Pre- and Post-Intervention.

Discussion

Improved opioid stewardship in the ED can contribute to combating the opioid epidemic. Our goal was not to prohibit the use of opioids in patients needing this therapy, but rather to guide prescribers to appropriate utilization. We have demonstrated that a CDST implemented within the EMR as well as direct feedback and coaching of ED practitioners leads to guideline compliant practices. A combination of active alerting through the EMR, passive guidance in assisting prescribers to act within regulatory limits, institutional- specific formularies, and direct transparent feedback to prescribers through email notification of monthly metrics through a dashboard and direct in-person feedback for outliers led to an overall improvement in guideline complaint opioid practices.

Other studies have examined the individual effects of utilizing EMR to reduce opioid overprescribing. We combined many of these methods to create our 4-tiered approach. Defaulting the dose, frequency, and quantity in the EMR has been studied previously and shown to be a promising approach to decreasing inappropriate opioid use and reducing practice variability. However, EMR defaults recognized limitation is that clinicians could unintentionally overprescribe or underprescribe for a particular complaint [16,17]. We understand that in our study, this also could be an unintended consequence of defaulting values but believe the benefit of limiting opioid defaults to 12 tablets outweighed the risk of over or underprescribing.

Additionally, creation of an opioid prescription policy has been shown to have a significant reduction in total opioid prescription and the number of pills per prescription in a smaller scale study [6]. We believe that standardizing our formulary improved overall prescribing practices in conjunction with our other interventions. One of the strengths of our intervention was the ability to provide direct feedback to clinicians through the use of an opioid dashboard. While kept in a separate location from the EMR, clinicians were emailed monthly with a link to the dashboard that was pushed directly to the clinicians. Some EMR vendors are in the process of integrating this information into their dashboards facilitating real time convenient feedback and access and we plan on moving ours in this direction as well. Other studies have supported this phenomenon of shared information demonstrating sharing of prescriber rates and engaging in feedback decreases opioid prescribing variability [14,15].

With the implementation of active EMR alerting, we sought to minimize alert fatigue. Consequently, we utilized three passive methods of decision support to complement the active alerts: (1) we removed certain medications from the ED formulary, (2) adjusted preset defaulted amounts in the opioid Medication orders, and (3) removed personal preference lists of opioid medications from the EMR. Using these other passive decision support methods minimized alert noise and served to encourage more appropriate prescribing practices. In a busy practice environment with competing demands, the path of least resistance for clinicians are EMR based defaults. If the EMR programming is incongruent with guidelines or recommendations, it is that much harder for clinicians to follow them. We recognized this as a barrier for clinicians and created the interventions listed in this paper to help guide compliant prescribing practices.

One limitation of our study is that we did not have a control group and therefore true attrition of prescribing changes is unknown. Additionally, our data did not allow us to assess how the state PDMP database may have influenced clinician prescribing preferences. However, a PDMP interface had already been embedded into the EMR prior to our intervention suggesting limited confounding. We were unable to separately assess this singular variable to determine how this may have influenced overall prescribing practices. Furthermore, we were unable to isolate how much of this reduction was from our intervention versus a general tendency nationally to limit opioid use. While successfully using our EMR, these interventions are subject to the limitations of individual EMRs. Additionally, since we implemented

Table 2

Pre- and post-intervention discharge Encounter data.

Discharge encounters with:

Pre-intervention

Post-intervention

Absolute % decrease

95% C.I.

N

Rate

N

Rate

Opioid prescriptions

53,337

14.4%

26,872

7.4%

7.0%

6.9%-7.2%

Opioid prescriptions exceeding 3 days’ Supply (12 tablets)

21,701

5.9%

2,352

0.7%

5.2%

5.1%-5.3%

Opioid prescriptions exceeding 30 MEDD

15,814

4.3%

1,130

0.3%

4.0%

3.9%-4.0%

Non-formulary opioid prescriptions

1,230

0.3%

311

0.1%

0.2%

0.2%-0.3%

Total discharge encounters

369,794

363,691

multiple interventions over a 6-month period, we were unable to address the efficacy of each individual intervention.

There may be instances where deviation from local guidelines is appropriate such as in the treatment of a hospice patient. Technical limitations precluded the analysis of appropriateness of these deviations.

In order to have adequate sampling for the clinician level analysis, we only included clinicians that had 25 discharges in both periods. While this may have led to undercounted data, all prescriptions were accounted for in the system level analysis.

Our practices were able to lead to overall and clinician level mindfulness in prescribing opioids from the ED. While our analysis was limited to the ED setting, similar measures can be easily applied

to other specialties. For example, creating standard formularies and defaults for post-operative pain management of specific procedures is feasible and in the process of being implemented. Likewise, creating specialty specific benchmarks and comparisons is ideal.

Conclusion

A combination of technical solutions and direct prescriber feedback is able to reduce the number, prescribed quantities, and types of opioid prescriptions written from the ED, in addition to assisting prescribers with meeting requirements for state-mandated regulations.

Fig. 2. Individual Providers Differences Pre- and Post-Intervention.

Grants/financial support

No outside funding provided support for this project.

Appendix A. Supplementary material

Supplementary data to this article can be found online at https://doi. org/10.1016/j.ajem.2019.158500.

References

  1. Centers for disease control and prevention (CDC). CDC grand rounds: prescription Drug overdoses–a US epidemic. MMWR Morb Mortal WklyRep, 2012; 61: p. 10-3.
  2. Guy GP, Zhang K, Bohm MK, et al. Vital signs: changes in opioid prescribing in the United States, 2006-2015. MMWR Morb Mortal Wkly Rep. 2017;66(26):697-704.
  3. Zgierska AE, Vidaver RM, Smith P, et al. Enhancing system-wide implementation of opioid prescribing guidelines in primary care: protocol for a stepped-wedge quality improvement project. BMC Health Services Res. 2018;18(1):415.
  4. Meisenberg BR, Grover J, Campbell C, Korpon D. Assessment of opioid prescribing practices before and after implementation of a health system intervention to reduce opioid overprescribing. JAMA Network Open. 2018;1(5) e182908.
  5. Hedegaard H, Warner M, Minino AM. Drug overdose deaths in the United States, 1999-2016. NCHS data brief, no 294. Hyattsville, MD: US Department of Health and Human Services, CDC, National Center for Health Statistics; 2017.
  6. Osborn SR, Yu J, Williams B, Vasilyadis M, Blackmore CC. Changes in provider prescribing patterns after implementation of an emergency department prescription opioid policy. J Emer Med. 2017;52(4):538-46.
  7. Axeen S, Seabury SA, Mechine M. Emergency department contribution to the prescription opioid epidemic. Ann Emerg Med. 2017;71(6):659-67.
  8. Hoppe JA, Kim H, Hard K. Associated of emergency department opioid initiation with recurrent opioid use. Ann Emerg Med. 2015;65:493-9.
  9. Barnett ML, Olenski AR, Jena AB. Opioid-prescribing patterns of emergency

    physicians and risk of long-term use. N Eng J Med. 2017;376:663-73.

    Jeffrey MM, Hooten M, Hess EP, et al. Opioid prescribing for opioid-naive patients in emergency departments and other settings: characteristics of prescriptions and association with long-term use. Ann Emerg Med. 2018;71:326-36.

  10. Delgado MK, Huang Y, Meisel Z, et al. National variation in opioid prescribing and risk of prolonged use for opioid-naive patients treated in the emergency department for Ankle sprains. Ann Emerg Med. 2018;72(4):389-400.e1.
  11. Zwank MD, Kennedy SM, Stuck LH, Gordon BD. Removing default dispense quantity from opioid prescriptions in the electronic medical record. Am J Emergency Med. 2017;35(10):1567-9.
  12. Chacko J, Greenstein J, Ardolic B, Berwald N. Effect of an emergency department opioid prescription policy on prescribing patterns. Am J Emergency Med. 2017;35 (9):1327-9.
  13. Burton JH, Hoppe JA, Echternach JM, Rodgers JM, Donato M. Quality improvement initiative to decrease variability of emergency physician opioid analgesic prescribing. Western J Emergency Med. 2016;17(3):258-63.
  14. Michael SS, Babu KM, Androski C, Reznek MA. Effect of a data-driven intervention on

    opioid prescribing intensity among emergency department providers: a randomized controlled trial. Academic Emergency Med: Official J Soc Acad Emergency Med. 2018;25(5):482-93.

    Zivin K, White JO, Chao S, et al. Implementing electronic health record default settings to reduce opioid overprescribing: A pilot study. Pain Med (Malden, Mass). 2019;20(1):103-12.

  15. Santistevan JR, Sharp BR, Hamedani AG, Fruhan S, Lee AW, Patterson BW. By default:

    the effect of prepopulated prescription quantities on opioid prescribing in the emergency department. Western J Emergency Med. 2018;19(2):392-7.

Leave a Reply

Your email address will not be published. Required fields are marked *